Dev open-sources GPT-2-tools to run original 1.5B GPT-2 XL locally on CPU
MikePFrank · x · 2026-10-02
Mike Frank released GPT-2-tools, a project that runs OpenAI's original 1.5B-parameter GPT-2 XL checkpoint locally on a CPU-only machine (13th-gen i7, 32GB RAM, WSL 2 Ubuntu), using Hugging Face's PyTorch/safetensors conversion of the historical checkpoint.
Key points:
- Ships complete PowerShell setup/run/chat scripts for one-command deployment under WSL 2
- The FP32 checkpoint fits in memory, though generation is much slower than on a modern discrete GPU
- Includes experiment scripts, tests, and verification/handoff docs
- The author notes the seed used for a prior demo output wasn't saved, so it can't be reproduced; future chats will save it
More from Infra
- awesome-jev indexes 700 production tools around TypeSafe AI's decision model Jev — Remarkable-Gur719 · 2026-10-02
- smolvm v1.22 ships near-instant VM resume for undoing agent actions, 6.5k stars — LoganGrasby · 2026-10-02
- Broadcom to lend Anthropic up to $42B for AI chips, eyeing top customer slot by 2027 — rohanpaul_ai · 2026-10-02
- Meta paper: only 50-60% of recommendation training time actually trained before optimizations — _reachsumit · 2026-10-02
- CoreWeave launches serverless GPUs: hourly-billed, no contract, private preview — altryne · 2026-10-02
- Japan plans $140B AI data center push; Huawei ships Mate90 with Taou chip — 创业邦 · 2026-10-02